11 MagneticEar Based Balancing ECCEAsia2011 01

7/25/2019 11 MagneticEar Based Balancing ECCEAsia2011 01

1/9

2011 IEEE

Proceedings of the 8th International Conference on Power Electronics (ECCE Asia 2011), The Shilla Jeju, Korea,May 30-June 3, 2011.

Magnetic Ear - Based Balancing of Magnetic Flux in High Power Medium Frequency Dual ActiveBridge Converter Transformer Cores

G. Ortiz

J. Mhlethaler

J.W. Kolar

This material is posted here with permission of the IEEE. Such permission of the IEEE does not in any way imply IEEEendorsement of any of ETH Zurichs products or services. Internal or personal use of this material is permitted. However,permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works forresale or redistribution must be obtained from the IEEE by writing to [email protected]. By choosing to view thisdocument, you agree to all provisions of the copyright laws protecting it.


2/9

8th International Conference on Power Electronics - ECCE Asia

May 30-June 3, 2011, The Shilla Jeju, Korea

978-1-61284-957-7/11/$26.00 2011 IEEE

[ThA1-3]

"Magnetic Ear"-Based Balancing of Magnetic Fluxin High Power Medium Frequency Dual Active

Bridge Converter Transformer Cores

G. Ortiz, J. Mhlethaler, and J. W. Kolar

Power Electronic Systems Laboratory, ETH Zurich

ETL I16, Physikstrasse 3

CH-8092 Zurich, Switzerland

Email: [email protected]

AbstractSemiconductor switches posses non-ideal behaviorwhich, in case of isolated DC-DC converters, can generate DC

voltage components in the voltage applied to the transformer.This DC voltage component is translated into a DC flux densitycomponent in the transformer core, increasing the risk ofdriving the core into saturation. In this paper, a novel non-invasive flux density measurement principle, called the "Mag-netic Ear", based on sharing of magnetic path between the mainand an auxiliary core is proposed. The active compensation ofthe transformer DC magnetization level using this transduceris experimentally verified. Additionally, a classification of thepreviously reported magnetic flux measurement and balancingconcepts is performed.

I. INTRODUCTION

Isolated and/or high step-up DC-DC converters are built

with arrangements of semiconductor switches which provideAC excitation to a transformer. Phenomena such as un-

matched turn-on/turn-off times, semiconductor forward volt-

age drop, gate driving signal delays or pulsating load, among

others, can cause differences in the positive and negative

volts-seconds applied to the transformer [1]. This results in

a DC voltage component at the transformer terminals, which

causes an undesired DC magnetic flux density component in

the transformer core.

To show the relation between this voltage and magnetic

flux density DC components, consider the circuit presented

in Fig. 1-a), where a Dual Active Bridge (DAB) DC-DC

converter topology is shown. The primary and secondary

bridges apply voltages vp(t) and vs(t) to the transformer re-spectively. The DC and the AC components of these voltagescan be separated into independent voltage sources, building

the circuit depicted in Fig. 1-b), where the secondary side

has been reflected to the primary side. Here, the resistances

Rp,T and R

s,T represent the winding resistances Rp,s and

Rs,s plus the semiconductors equivalent on-state resistancesof the primary and secondary side switches respectively.

In steady state, the DC magnetizing current Im,DC of thetransformer is given by

Im,DC=Ip,DC I

s,DC=Vp,DC

Rp,T

Vs,DC

Rs,T

. (1)

The DC magnetic flux density is then determined by the

characteristics of the winding and core through

LmRp s, Rs s,

a)

ip( )t is( )t

Primary Bridge Secondary Bridge

vp( )t vs( )t

1:n

im( )t

c)

Rp T, Lm

Vp DC, Vs DC,

vs AC,

Rs T,

is( )tim( )tip( )t

vp AC,b)

Ferrite coreCopper litz

wire

Air-cooled

heat-sink

TeflonIsolation

Fig. 1. a) DAB converter with simplified transformer model;b) Equivalent model of the converter with independent DC andAC voltage sources and reflected secondary side; c) Shell-typetransformer optimized for efficiency

BDC= Im,DC Np

lm0r

=

Vp,DC

Rp,T

Vs,DC

Rs,T

Np

lm 0r,

(2)

where lmis the length of the magnetic path, Npis the numberof turns in the primary side, 0 is the permeability if air and

r is the cores relative permeability in the linear region of

the B-H curve.From (2) it can be seen that the DC magnetic flux density

is limited by the equivalent series resistances,Rp,T andR

s,T,


3/9

Flux Measurement

Saturation Detection Dynamic Flux Meas. Continuous Flux Meas.

Flux Balancing

Passive Control Saturation Correction Active Flux Control

Parallel flux path [2-3] RC integration [4] Mag. current [5-7] Se ries capacitor Volts-seconds

correction [2]

Volts-sec. continuous

[5,7,9,12]correction

Air gap

Device matching

Unipolar flux

Orthogonal flux [6]

Current meas. [9]

Hall sensing [10]

Proposed MethodMagnetic Ear

Proposed MethodMagnetic Ear

Fig. 2. Classification of previously proposed flux balancing concepts. The two main areas are flux measurement and flux feedback control.The proposed measurement concept, the "Magnetic Ear", is highlighted within this classification.

of the circuit, which are typically kept as low as possible

in order to decrease the converters conduction losses. Thismeans that a small DC component in the voltage applied tothe transformer generates a large DC flux density component.

For example, taking the shell-type 166 kW/20kHzefficiency-optimized transformer (cf. Fig. 1-c)) with suitable

switches on the primary side, the primary side equivalent

resistance Rp,T reaches 1.7 m. This design considers aFerrite N87 core material which is characterized by a relative

permeabilityr around1950 . In this design, a0.00125 %ofrelative difference in the duration of the positive and negative

semicycles of the primary voltagevs(t), i.e. a switching timeerror of 0.25ns, would suffice to create a DC flux densitycomponent ofBDC= 50mT.

With this DC flux density component, the core can be eas-ily driven outside the linear region of the B-H curve, generat-

ing a non-linear magnetizing current with high peak values.This results in increased conduction and switching losses,

causing reduction in efficiency and higher semiconductor and

transformer operating temperatures which could ultimately

destroy the converter. Moreover, a DC biased flux density

waveform results in higher core losses, further compromising

the converters efficiency. For these reasons, the operation of

the transformer under balanced conditions, i.e. with zero DC

flux density component, must be ensured. It is also worth to

note that if a balanced operation of the flux density in the

core is ensured, the transformer can be designed with low

flux density over-dimensioning, meaning that its magneticcross section, and therefore its volume, can be reduced,

increasing the converters power density.

In this paper, a novel magnetic flux density transducer

is introduced and its operating principle is experimentally

verified. The measured transducer signal is used to perform

a closed-loop flux balancing control, ensuring the operation

of the transformer core within safe flux density values. First,

inSection II, a classification of previously proposed methods

which enable detection of saturation and/or continuous mea-

surement of the flux density behavior is laid out along with

methods to balance the transformers internal flux density.

The proposed flux density transducer is described in Sec-

tion IIItogether with experimental testing. InSection IVtheoutput signal of the transducer is used to perform a closed-

loop compensation of the flux density in the transformer core.

vm 2, ( )tvm 1, ( )t

vm( )t

vm( )t

VDC p, vp( )t

vp( )t

vs( )t

vp( )t

vs( )t

a) b)

c)

Fig. 3. Previously proposed concepts for magnetic saturationprevention: a) Parallel magnetic path in an E-core with an gapedleg [2]; b) Parallel magnetic path with external cores and reducedcross section [3], c) Integration of applied voltage with RC network[4].

II . CLASSIFICATION OF F LU X M EASURMENT/C ONTROL

METHODS

In order to ensure balanced flux operation, the mainproblems that must be addressed are A. measurement of

the cores internal flux status and B. balancing or closed-

loop control of the flux. Within these two topics, other

sub-categorizations are possible, as displayed in Fig. 2 and

discussed in the following sections.

A. Flux Measurement / Saturation Detection

The methods for recognition of the cores flux state can

be classified into 1) saturation detection, 2) dynamic flux

measurement and 3) continuous flux behavior measurement.

The measurement method proposed in this paper lays in thislast class.

1) Saturation Detection: In [2], an E-core was used withan air-gap in one of the external legs (Fig. 3-a)). During

normal operation, the flux flows only in the un-gapped leg


4/9

but as soon as this leg saturates, a flux is forced to the

gapped leg and therefore a voltage can be induced in an

additional winding, detecting the saturation of the main flux

path. Alternatively, in [3] a slot is placed in one of the core

legs, as shown in Fig. 3-b) in order to reduce the cross

section area in this place. An additional magnetic path with

a winding is provided parallel to the slotted part of the core.

As the slotted section has a smaller area, it saturates before

the rest of the core and magnetic flux is forced into the

parallel magnetic path. This induces a voltage in a winding

indicating the impending core saturation.

With both these methods, only the saturation of the core

is detected, which may be enough in some applications. In

applications which require high efficiency, however, this is

not enough since the flux density in the core can still be

biased without being saturated and therefore the core losses

are increased. Moreover, to implement both this methods,

modifications to the magnetic components are required,

increasing costs and complexity.

2) Dynamic Flux Measurement: This method was pro-

posed in [4] to detect flux unbalance due to variations in the

converter loading conditions. The principle is to perform an

integration of the applied voltage through an RC network

or an active integrator (cf. Fig. 3-c)). This integrated signal

is proportional to the cores magnetic flux. Due to the

requirement of an integration, this method can only sense

dynamic variations of the flux, i.e. steady state asymmetries

will not be detected.

3) Continuous Flux Behavior Measurement: The sens-

ing of the flux behavior with large bandwidth and indepen-

dent of the operating conditions has been covered by severalpublications, where the following main categories can be

identified:

a) Magnetizing Current Measurement: The magnetiz-

ing current im(t) indicates the status of flux density in the

vm( )t

vm( )t

vm( )tvp( )t

vs( )t

a) b)

c)

Fig. 4. Continuous measurement of cores internal flux: a) Con-struction of magnetizing current with external transformer [5],b) Magnetic flux measurement with hall sensor in magnetic path;c) Orthogonal magnetic fluxes [6].

core. The measurement of this current through substraction

of the scaled primary and secondary currents was proposed

in [4].

In [5] and later in [7], a measurement of the magnetizing

current was performed by building an additional transformer

with the same turns ratio as the main transformer but withone of the windings in the inverted orientation (cf. Fig. 4-

a). As the primary and secondary currents flow through this

transformer, most of the magnetic flux is canceled out and

the remaining flux, which is proportional to the magnetizing

current, is measured with an additional winding.

The disadvantage of this method is the requirement of

isolation on the additional transformer, which needs to be at

least the same as the one of the main core. Also, practical

issues may arise in higher power transformers where the

wiring of primary and secondary sides has an increased

complexity.

b) Orthogonal Magnetic Fluxes: In [6] the internalcore flux was measured by using an additional coil fed by

a DC current which generates a magnetic flux orthogonal

to the main flux (cf. Fig. 4-c)). The orthogonality of the

magnetic fluxes ensures that no voltage is induced in the

additional coil due to the main flux. As the main magneticflux density is changed, the B-H characteristic of the material

is also changing. This material property change is translated

in a variation of the flux in the orthogonal coil, inducing a

voltage in its terminals. This principle was also proposed for

microfabricated inductors [8] to intentionally shape the B-H

loop of the magnetic material. In this concept, modified or

specially shaped cores are required to insert the orthogonal

winding, increasing its cost and complexity. Moreover, if theB-H loop has a large linear zone, voltage would be induced

in the orthogonal coil only when the core is saturated.

c) Converter Current Measurement and Processing:

The direct measurement of the primary and/or the secondary

currents has also been used to balance the flux in the

core. As an example, the DC magnetization of the core

generates primary/secondary currents with even numbered

Fourier components. The amplitude of these components can

be measured and used as feedback signal to balance the

transformer flux, as was performed in [9].

In converters with modulations which do not operate

always at 50% duty cycle, only the magnetizing current ispresent during the freewheeling periods. This current can bemeasured during these intervals obtaining information about

the status of the cores flux.

d) Flux Observer: To overcome the limitation of only

dynamic flux measurement, the method described in Sec-

tion II-A2 can be complemented with a measurement of

the transformer current [4]. This way, an observer that

reconstructs the flux density based on these two signals can

be implemented.

e) Hall Sensing: The most direct way to measure the

flux in the core would be to insert a thin hall sensor in the

magnetic flux path [10] (cf. Fig. 4-b). However, this requires

the insertion of an air gap in the magnetic core, reducing themagnetizing inductance and increasing the reactive power

provided to the transformer.


5/9

Ls Ws

Rm

vp( )t

Auxiliarycore + winding

Main core

a)

b) c)

Vp Vsvs( )t

Aux.circ.

Fig. 5. a) Proposed flux density measurement concept with auxiliarycore and winding sharing magnetic path with the main core. Theauxiliary core can be placed so that its flux is b) parallel or c)orthogonal to the main flux. The auxiliary core is not shown toscale.

B. Flux Balancing / Feedback Control

The internal core flux can be passively or actively bal-

anced. Depending on whether the measuring principle detects

core saturation or performs a complete flux measurement,

the active flux balancing principles can be subdivided intosaturation correction or continuous flux control.

1) Passive Balancing: Passive balancing refers to any

balancing principle which does not modify the switching

behavior of the semiconductor devices in order to keep the

transformer flux between safe margins. The following passive

flux balancing principles can be pointed out:

a) Series Capacitor: One of the most utilized flux

balancing principles, due to its simplicity, is the inclusion of

a capacitor in series to the transformer winding. The main

disadvantages of this approach are (1) increased converter

volume, (2) increased converter losses and (3) slow dynamic

response. This idea was further developed in [11] where a

resistor was placed in parallel to the capacitor in order toimprove low-frequency behavior.

b) Air-gap in Cores Magnetic Path: When an air-gap

is included in the cores magnetic path, the permeability of

the core is effectively decreased. This in turn increases the

tolerable DC magnetization but doesnt eliminate the DC flux

component as with a series capacitor, thus this is not strictly

a flux balancing method. Moreover, the inclusion of an air-

gap in the core decreases the value of magnetizing inductance

Lm, increasing the switched and conducted currents.

In addition to the previously presented passive flux bal-

ancing principles, the prevention of core saturation can be

included as part of the converter design process. In [1] the

different converter electrical parameters that influence thecore saturation were clearly pointed out and used to give

design considerations which help avoiding it.

2) Active Saturation Correction: The flux measuring

concepts presented in Section II-A1 can be used to imple-

ment feedback control which only operates under impending

core saturation, as was done in [2].

3) Active Feedback Control of the Flux: If a signal

proportional to the internal core flux density is available,

the DC magnetization of the core can be actively controlled

by modifying the volts-seconds applied to the transformer.

The main feedback flux control principles that have been

proposed are detailed in [46, 10, 12]. The feedback scheme

proposed in this paper is revised in Section IV.

III. PROPOSED F LU X M EASUREMENT M ETHOD

This section introduces the proposed flux density trans-

ducer. First, the working principle is explained and af-

terwards this principle is experimentally verified using an

adequate test circuit.

A. Working PrincipleThe proposed flux measurement concept is explained

through Fig. 5-a). Here, the main core and an auxiliary core

with a winding Ws and an inductanceLs are displayed. Thisadditional core is attached to the main core so that they share

a part of the magnetic path, represented by the reluctance

Rm in Fig. 5-a). The value of the inductance Ls measuredat the terminals of the auxiliary winding Ws decreases withincreasing value of reluctance Rm. This reluctance in turnis inversely proportional to the relative permeability r ofthe main core. Therefore, as the core is driven closer to

saturation, the value ofr is decreased, increasing Rm andfinally decreasing the measured inductance Ls. This drop in

inductance would thus directly indicate that the main core isentering the saturation region.

This principle was tested in two different cores: a gapped

E55 N27 E ferrite core and a AMCC80 Amorphous C-cut

core. The auxiliary core is half of a E25 N87 core with

an auxiliary winding consisting on 14 turns. It should be

noted that the auxiliary core can be placed on the surface

of the main core so that their magnetic fluxes are parallel

(Fig. 5-b)) or, in order to avoid induced voltages in the

auxiliary winding, orthogonal to each other (Fig. 5-c)). The

measured inductance for different magnetization levels is

shown in Fig. 6-a) and b) for the Ferrite and AMCC80 cores

respectively. Additionally, the positive part of the respective

B-H loops is displayed. In case of the ferrite core, a cleardrop in the measured inductance with both parallel and

orthogonal placements can be seen around 0.35 T wherethe core enters saturation region. In case of the AMCC80

core, as shown in the upper part of Fig. 6-b), its B-H loop

is characterized by a continuous decrease of permeability,thus a smooth drop in inductance is measured as the cores

magnetic field is increased.

B. The Magnetic Flux Density Transducer:

The "Magnetic Ear"

In order to measure the inductance seen from the terminals

ofWs (cf. Fig. 5-a)), an auxiliary circuit is required. The

utilized circuit in this case is shown in Fig. 7. Here, a half-bridge applies a square-shaped voltage vaux(t) excitationwith high-frequency to the auxiliary winding. This frequency


6/9

FluxDensity[T] 0.5

0.1

0.2

0.3

0.4

0

0 2 2.5 3 3.5 4

Inductance[H]

160

80

100

120

140

60

FluxDensity[T] 1.5

0.5

1

0

Inductance[H]

55

40

45

50

35

Magnetic Field [kA/m]a) b)

0 0.2 0.4 0.6 0.8 11.51Magnetic Field [kA/m]

Ls parallel,

Ls orthogonal,Ls orthogonal,

Ls parallel,

Voltage[V]

Voltage[V]

Time [ms]c)

20.5 1 1.500

2

1

0.5

1.5

-150

150

0

-75

75

Current[A]

-15

15

0

-7.5

7.5

i tm( )

v tp( )

v tm( )

Fig. 6. B-H loops and inductance measurement results on a) ferrite core b) AMCC80 Metglas core. In the upper part of c), a square-shapedvoltage applied to the mentioned ferrite core drives the core to saturation. The change in permeability is shown in the lower part of c),

which was sensed using the circuit presented in Fig. 7-a).

VIN

vm( )t

vaux( )t

iaux( )t

RrCr

R1

R2

iaux( )t

nr

1:nrMainCore

Magnetic earand auxilary circuit

a)

b) Supply inputVin

Half-bridge

Auxiliary core

Output signal

v tm( )

Rectifier

Currenttransformer

Fig. 7. a)The magnetic ears driving auxiliary circuit. A half-bridgeapplies a square-shaped voltage with high frequency to the magneticear winding. The current through this core is measured and rectifiedto obtain a voltage inversely proportional to the inductance of theauxiliary core; b) The Magnetic Ear hardware realization.

needs to be several times higher as the main core excitation.The applied square shaped voltage generates a triangular

shaped current iaux(t) through the winding, whose peak

value is inversely proportional to the auxiliary inductance

value. This current is sensed with a current transformer

and rectified with at full-wave diode rectifier. The rectified

current is fed to an RC network for low-pass filtering andfinally amplified to obtain signal vm(t) which is inverselyproportional to the auxiliary inductanceLs.

The magnetic flux density transducer, introduced here as

the "Magnetic Ear", was tested using the circuit shown in

Fig. 8. Here, a single full-bridge converter applies square-

shaped voltage with peak value Vp. This way, the magne-

tizing current im(t) can be directly measured and analyzedto study the state of the magnetic flux density in the core.

Moreover, an additional winding carrying a DC currentId isplaced around the main core in order to intentionally force

a DC flux density component in the core.

The results of the test are shown in Fig. 6-c). Here, a 1 kHz

square-shaped voltage is applied to the winding and, as can

be noticed from the peaks in the magnetizing current im(t),the core is driven into saturation. Additionally, a DC bias

in the flux density is observed since the magnetizing current

has a non-symmetric shape. The output of the Magnetic Ear

is shown in the lower part of Fig. 6-c) where high peaks in

the output signal vm(t) are detected when the core is taken

to saturation. It should be noted that the output signal of the

vp( )t

Main core

Vp

Aux.circ.

Magnetic Ear

im( )t Vd

Ld Id

Fig. 8. Test circuit used to analyze the performance of the MagneticEar. A single full-bridge converter is used to magnetize the core andan external circuit is used to impose a DC flux density in the core.


7/9

SOC1

SOC2

BDC m,

MAF Look-up

vm( )t

BDC*

CB

D

Full-bridge

D

SOC1

SOC2

vp( )tim( )t

Ts 2Ts

M

ainCore

Aux

.

circ

.

PWMmod.

0

vp( )t

vm( )t

4

a) b)

im( )t

1/2

Fig. 9. a) Feedback/balancing scheme: the output of the magnetic ear,vm(t) is sampled by the ADC in the DSP board. By sampling oneach switching event of the full-bridge (part b)), a construction of the DC flux density in the core is obtained and used to balance theflux in the main core.

Magnetic Ear is higher when the full-bridge switches frompositive to negative as it is when switching from negative

to positive. This shows that the flux density contains a DC

component and also gives information about the polarity of

this DC bias. The means to use this signal to compensate the

DC bias in the core will be now discussed

IV. CLOSEDL OO P C ONTROL OF THEDC FLU X D ENSITYC OMPONENT

The output vm(t) of the Magnetic Ear must be fed intothe DSP controller of the full-bridge converter in order to

actively compensate the DC-bias. To obtain the value of

BDC in the main core, the scheme shown in Fig. 9-a)

was used. As can be seen from Fig. 9-b), different Startof Conversion (SOC) signals (the signals that trigger a

conversion of the analog-to-digital converter of the DSP), i.e.

SOC1 and SOC2 are generated in the positive and negative

edges of the full-bridge output voltage vp(t). The sampledvalues at each of these instants are independently stored and

filtered by Moving Average Filters (MAF). To obtain the

final flux density value, a look-up table is built based on the

measurements of the Magnetic Ear output and the magnetic

flux density calculated from the voltage applied to the core. It

should be noted that the gain from the Magnetic Ear output to

flux density is highly non-linear since, when the core is close

to saturation, a small increase in the flux density generates a

large change in the output signal of the transducer whereas,when the core is in linear region, the change in output signal

with respect to changes in the flux density is several times

smaller.

The output of the look-up tables are the absolute values

of flux density in the main core at the switching instants.

Therefore, the substraction of these two signals gives twice

the value of DC flux density component BDC inside core.For example, in steady state operation, if no DC bias is

present in the core, the output of the Magnetic Ear would

be identical at the positive and negative edges of the applied

voltage vp(t). As a consequence, the output of the look-up tables, the flux density at the switching times, would be

identical and thus the measured DC flux density componentwould be zero.

To control the DC flux density in the core, a standard

Time [ms]

0

-75

-150

75

150

0 210.5 1.50

1

0.5

0.25

0.75

210.5 1.5Time [ms] Time [ms]

-100

100

0

-50

50

Voltage[V]

Voltage[V]

DCFluxDensity[m

T]

0

-10

10

0

-5

5

Current[A]

a)

b) c)

0 20 2510 155 17.5 22.57.5 12.52.5

v tm( )v tm( )

v tp( )

i tm( )i tm( )

B tDC( )

B tDC( )*

v tp( )

Fig. 10. Closed loop results for step response in the DC flux densityreference BDC: a) Response of the measured flux density to anincrease in reference from 75mT to 75mT. The behavior ofthe applied voltage vp(t), the magnetizing current im(t) and theMagnetic Ear output signal vm(t) b) before and c) after the stepapplication is also shown.

PI controller, CB is used (cf. Fig. 9-a)). The output ofthis controller is the additional duty cycle D required toincrease or decrease the DC voltage applied to the primary

winding. In the PWM modulator, this signal is combined

with the duty cycle D calculated to transfer desired amountof power, which was left to 50 % in this case.

Two tests were performed to analyze the performance of

the closed loop compensation method: A. step response and

B. disturbance compensation. These tests were performed

using the ferrite core with a1.25kHz square shaped voltagewhit peak value of50 V.

A. Step ResponseIn Fig. 10-a), the response to a step in the DC flux

reference value (keeping the disturbance current Id of the


8/9

0Time [ms]

20 2510 155 17.5 22.57.5 12.52.5

Time [ms] Time [ms]

-10

10

0

-5

5

Current[A]

a)

b)

-50

-100

-150

0

50

-100

100

0

-50

50

Voltage[V]

Voltage[V]

DCFl

uxDensity[mT]

c)

0 210.5 1.5 210.5 1.50

v tm( )v tm( )

v tp( )

i tm( )i tm( )

B tDC( )

B tDC( )*

v tp( )

0

1

0.5

0.25

0.75

IdId

Fig. 11. Closed loop results for disturbance compensation: a) Re-sponse of the measured flux density to an externally applied DCflux density of 130mT; The behavior of the applied voltagevp(t), the magnetizing current im(t) and the Magnetic Ear outputsignal vm(t) b) before and c) after the activation of the feedbackcompensation.

external circuit, cf. Fig. 8, at zero) from 75mT to 75mTis displayed. As can be seen, the flux density reference is

followed closely by the measurement. The behavior of the

magnetizing current im(t) and the output of the MagneticEarvm(t) before and after the step application are displayedin Fig. 10-b) and -c) respectively. Before the step application,

the magnetizing current im(t) features a negative DC bias,which is generating the 75mT set as reference. After thestep application, the DC value of this current has a positiveDC bias, meaning that the DC flux density component has

changed its polarity. It is worth to note that, before the step

application, the output signal vm(t) of the Magnetic Earhas its peaks in at the positive edges of the applied voltage

vp(t), meaning that at these points the highest flux densityvalue is encountered. After the application of the step, the

peaks appear at the negative edges of the applied voltage, i.e.

180 phase-shift with respect to the starting condition, whichshows that the DC flux density has changed its polarity.

B. Disturbance Compensation

Using the external circuit shown in Fig. 8, a DC flux

density of130mTcomponent is imposed in the main coreto test the performance of the feedback compensation to

external disturbances. To do this test, first the external distur-

bance is imposed in the core without feedback compensation

of the flux density. Later, the compensation is activated.

The results of this test are shown in Fig. 11-a), where

the disturbance current Id reflected to the primary side(the full-bridge output side) is shown. The compensation is

turned on around t=5 ms and, as can be seen, the feedbackcompensation scheme is able to bring the DC flux densitycomponent back to zero in spite the externally forced DC

flux density. The magnetizing current before the disturbance

application, shown in Fig. 11-b), features no DC component

but, however, the output of the Magnetic Ear has clear peaks

at the positive edges of the the full-bridge output voltage,

meaning that the DC flux density component in the core

is negative. After the feedback compensation is activated

(cf. Fig. 11-c)), the magnetizing current im(t) shows a DCcomponent which compensates for the externally applied

DC flux, which is the inverted value of current Id. Theoutput signal from the Magnetic Ear has equal peaks at the

switching events, which shows that the flux density operates

under balanced conditions.

C. Comments on Compensation Scheme Bandwidth

As shown in Figs. 10 and 11, the time constant of the

closed loop compensation scheme is around 8 ms, which isequivalent to 10 switching cycles of the full-bridge con-verter. This time can be reduced by implementing a feed-

forward scheme which uses the measured DC flux densityBDC signal, since the required volts-seconds required tocompensate a given DC bias in the flux density can be easily

calculated. Thus, a cycle-per-cycle compensation could be

implemented to avoid DC magnetization due to, for example,

loading condition variations. Other techniques such as a

dead-beat control could be implemented in order to increase

the bandwidth of the compensation loop.

Since the switching frequency in medium-frequency appli-

cations, where the Magnetic Ear would provide considerable

benefits, is relatively low, a sampling frequency several times

higher than the main converters switching frequency could

be achieved. This way, a continuous tracking of the flux

density would be possible and thus a faster compensationloop would be feasible.

V. CONCLUSIONS

A new magnetic flux density transducer, the Magnetic

Ear, has been introduced in this paper. The basic working

principle of this transducer is the sharing of magnetic path

between the main core and an auxiliary one. The influence

of the main flux in this shared magnetic path is sensed

by measuring the inductance in an additional winding in

the auxiliary core. An auxiliary circuit is used to obtain

a signal inversely proportional to the inductance in the

auxiliary winding. This signal is fed to a DSP which,

based on sampling at the switching events of the drivingfull-bridge, constructs the signal proportional to the DC

flux density component inside the core. With this signal, a

closed loop compensation was successfully tested under step

changes in the DC component reference and external DC flux

disturbances, thus the proposed concept can be implemented

in isolated DC-DC converter to avoid DC magnetization ofthe transformer core.

REFERENCES

[1] K. OMeara, Passive Balancing of Transformer Flux in Power Con-verters, in Proc. of the 10th National Solid-State Power ConversionConference (POWERCON), vol. A1, March 1983, pp. 111.

[2] R. Patel, Detecting Impending Core Saturation in Switched-ModePower Converters, in Proc. of the 7th National Solid-State PowerConversion Conference (POWERCON), vol. B3, March 1980, pp. 111.


9/9

[3] S. Klopper and J. A. Ferreira, A Sensor for Balancing Flux in Con-verters with a High-Frequency Transformer Link, IEEE Transactionson Industry Applications, vol. 33, pp. 774779, May/June 1997.

[4] D. Wilson, A New Pulsewidth Modulation Method Inherently Main-tains Output Transformer Flux Balance, in Proc. of the 8th NationalSolid-State Power Conversion Conference (POWERCON), vol. D1,

April 1981, pp. 114.[5] F. Stgerer, J. W. Kolar, and U. Drofenik, A Novel Concept for

Transformer Volt Second Balancing of VIENNA Rectifier III Basedon Direct Magnetizing Current Measurement, in Proc. of the NordicWorkshop on Power and Industrial Electronics Workshop, June 2000,pp. 134139.

[6] W. M. Polivka, A. Cocconi, and S. Cuk, Detection of MagneticSaturation in Switching Converters, in Proc. of the PCI Conference,March 1982, pp. 584597.

[7] A. Gerstman and S. Ben-Yaakov, Zeroing Transformers DC Currentin Resonant Converters with no Series Capacitors, in Proc. of the

Energy Conversion Congress and Exposition, September 2010, pp.40284034.

[8] L. Daniel, C. R. Sullivan, and S. R. Sanders, Design of Microfab-ricated Inductors, IEEE Transactions on Power Electronics, vol. 14,pp. 709723, July 1999.

[9] R. Kuttner, Circuit Techniques for Eliminating Primary Current Un-

balance in Push-Pull Power Converters, in Proc. of the 7th NationalSolid-State Power Conversion Conference (POWERCON), vol. F2,March 1980, pp. 19.

[10] F. P. Dawson, DC-DC Converter Interphase Transformer DesignConsiderations: Volt-Seconds Balancing,IEEE Transactions on Mag-netics, vol. 26, pp. 22502252, September 1990.

[11] R. Redl, N. Sokal, and C. W. Schaefer, Transformer Saturation andUnusual System Oscillations in Capacitively Coupled Half-Bridge orFull-Bridge Forward Converters: Causes, Analyses, and Cures, inProc. of the Power Electronic Specialists Conference (PESC) , April1988, pp. 820828.

[12] S. Han, I. Munuswamy, and D. Divan, Preventing TransformerSaturation in Bi-Directional Dual Active Bridge Buck-Boost DC-DC Converters, in Proc. of the Energy Conversion Congress and

Exposition, September 2010, pp. 14501451.

Documents

11 MagneticEar Based Balancing ECCEAsia2011 01